Image processing system for region-of-interest-based video compression

Abstract

An apparatus for remote processing of raw image data receives the raw image data from a camera, such as a security camera. The apparatus includes a detection module to detect portions of the image data that contain possible regions of interest. Information indicating the portions that contain the possible regions of interest is then used during a compression process so that the portions that contain the possible regions of interest are compressed using one or more compression algorithms to facilitate further analysis and the remainder are treated differently. The compressed image data is then transmitted to a central system for decompression and further analysis. In some cases, the detection system may detect possible regions of interest which appear to be faces, but without performing full facial recognition. These parts of the image data are then compressed in such a way as to maintain as much facial detail as possible, so as to facilitate the facial recognition when it is carried out at the central server. The detection may be performed on the raw image data or may be performed as part of the compression process after a transformation of the raw image data has been carried out.

The invention claimed is: 
     
       1. A system comprising:
 a first non-transitory memory; 
 one or more first hardware processors coupled to the first non-transitory memory and configured to read first instructions from the first non-transitory memory to cause the one or more first hardware processors to:
 receive raw image data in a raw form from a source of raw image data; 
 perform a first domain transformation, selected from a predetermined set of domain transforms, on the raw image data; 
 detect one or more portions of the raw image data containing one or more possible regions of interest based on the domain transformed raw image data; 
 select a first compression process from a plurality of predetermined compression processes based at least in part on a data type associated with the raw image data, wherein the data type indicates whether the raw image data represents a frame of video; 
 compress the raw image data, in each of the detected portions based on the domain transformed raw image data, at a first compression level using the first compression process to produce compressed detected data; and 
 treat remainder portions of the raw image data that do not form part of the detected portions either by compressing the remainder portions at a second compression level different than the first compression level or by discarding the remainder portions; and 
 
 a central computing system configured to identify one or more elements of interest in the compressed detected data. 
 
     
     
       2. The system of  claim 1 , wherein compressing the raw image data at the first compression level facilitates further analysis of the compressed detected data when the compressed detected data has been decompressed. 
     
     
       3. The system of  claim 1 ,
 wherein the first compression process is selected so that the compressing of the detected portions at the first compression level facilitates further analysis of the compressed detected data when the compressed detected data has been decompressed. 
 
     
     
       4. The system of  claim 3 , wherein execution of the first instructions further causes the one or more first hardware processors to:
 determine that the remainder portions of the raw image data are to be compressed; and 
 select a second compression process from the plurality of predetermined compression processes to be used to compress the remainder portions of the raw image data to produce compressed remainder data. 
 
     
     
       5. The system of  claim 1 , wherein execution of the first instructions further causes the one or more first hardware processors to:
 detect possible facial regions of interest which appear to be faces, but without performing full facial recognition; and 
 compress the detected possible facial regions of interest using a compression process that maintains as much facial detail as possible, so as to facilitate full facial recognition carried out by the central computing system. 
 
     
     
       6. The system of  claim 1 , further comprising a camera, wherein the camera includes the source of the raw image data. 
     
     
       7. The system of  claim 1 , wherein the central computing system comprises:
 a second non-transitory memory; and 
 one or more second hardware processors coupled to the second non-transitory memory and configured to read second instructions from the second non-transitory memory to cause the one or more second hardware processors to:
 receive at least the compressed detected data from the one or more first hardware processors; 
 decompress at least the compressed detected data; 
 analyze the one or more possible regions of interest in the decompressed detected data to determine whether the one or more possible regions of interest contain the one or more elements of interest; and 
 identify at least one of the one or more elements of interest. 
 
 
     
     
       8. The system of  claim 7 , wherein the one or more possible regions of interest are one or more possible facial regions of interest that may contain one or more faces, the one or more elements of interest are one or more faces, and wherein execution of the second instructions further causes the one or more second hardware processors to identify the one or more faces of the one or more elements of interest by performing full facial recognition. 
     
     
       9. The system of  claim 1 , wherein the performing of the first domain transformation on the raw image data includes performing a Haar transformation to generate Haar coefficients, and wherein the detecting of the one or more portions of the raw image data is based on the Haar coefficients. 
     
     
       10. The system of  claim 1 , wherein the compressing of the raw image data in each of the detected portions comprises performing a second domain transformation on the raw image data. 
     
     
       11. A method comprising:
 receiving raw image data in a raw form from a source of raw image data; 
 performing a first domain transformation, selected from a predetermined set of domain transforms, on the raw image data; 
 detecting one or more portions of the raw image data containing one or more possible regions of interest based on the domain transformed raw image data; 
 selecting a first compression process from a plurality of predetermined compression processes based at least in part on a data type associated with the raw image data, wherein the data type indicates whether the raw image data represents a frame of video; 
 compressing the raw image data, in each of the detected portions based on the domain transformed raw image data, at a first compression level using the first compression process to produce compressed detected data; 
 treating remainder portions of the raw image data that do not form part of the detected portions either by compressing the remainder portions at a second compression level different than the first compression level or by discarding the remainder portions; and 
 transmitting at least the compressed detected data to a central computing system configured to identify one or more elements of interest in the compressed detected data. 
 
     
     
       12. The method of  claim 11 , wherein compressing the raw image data at the first compression level facilitates further analysis of the compressed detected data when the compressed detected data has been decompressed. 
     
     
       13. The method of  claim 11 , wherein the first compression process is selected so that the compressing of the detected portions at the first compression level facilitates further analysis of the compressed detected data when the compressed detected data has been decompressed. 
     
     
       14. The method of  claim 13 , wherein the treating of the remainder portions comprises:
 determining that the remainder portions of the raw image data are to be compressed; and 
 selecting a second compression process from the plurality of predetermined compression processes to be used to compress the remainder portions of the raw image data to produce compressed remainder data. 
 
     
     
       15. The method of  claim 11 , further comprising:
 detecting possible facial regions of interest which appear to be faces, but without performing full facial recognition; and 
 compressing the detected possible facial regions of interest using a compression process that maintains as much facial detail as possible, so as to facilitate full facial recognition carried out by the central computing system. 
 
     
     
       16. The method of  claim 11 , wherein a camera includes the source of the raw image data. 
     
     
       17. The method of  claim 11 , further comprising, at the central computing system:
 receiving the at least the compressed detected data; 
 decompressing the at least the compressed detected data; 
 analyzing the one or more possible regions of interest in the decompressed detected data to determine whether the one or more possible regions of interest contain the one or more elements of interest; and 
 identifying at least one of the one or more elements of interest. 
 
     
     
       18. The method of  claim 17 , wherein the one or more possible regions of interest are one or more possible facial regions of interest that may contain one or more faces, the one or more elements of interest are one or more faces, and wherein the identifying of the at least one of the one or more elements of interest comprises identifying the one or more faces of the one or more elements of interest by performing full facial recognition. 
     
     
       19. The method of  claim 11 , wherein the performing of the first domain transformation on the raw image data includes performing a Haar transformation to generate Haar coefficients, and wherein the detecting of the one or more portions of the raw image data is based on the Haar coefficients. 
     
     
       20. The method of  claim 11 , wherein the compressing of the raw image data in each of the detected portions comprises performing a second domain transformation on the raw image data.